Two effects are mainly responsible for the observed enthalpy change in protein unfolding: the disruption of internal interactions within the protein molecule (van der Waals, hydrogen bonds, etc.) and the hydration of groups that are buried in the native state and become exposed to the solvent upon unfolding. In the traditional thermodynamic analysis, the effects of hydration have usually been evaluated using the thermodynamic data for the transfer of small model compounds from gas phase to water. The contribution of internal interactions, on the other hand, are usually estimated by subtracting the hydration effects from the experimental enthalpy of unfolding. The main drawback of this approach is that the enthalpic contributions of hydration, and those due to the disruption of internal interactions are more than one order of magnitude larger than the experimental enthalpy value. The enthalpy contributions of hydration and disruption of internal interactions have opposite signs and cancel almost completely, resulting in a final value that is over ten times smaller than the original terms. For this reason the classical approach cannot be used to predict unfolding enthalpies from structure: any error in the estimation of the hydration enthalpy will be amplified by a factor of ten or more in the estimation of the unfolding enthalpy. Recently it has been shown that simple parametric equations which relate the enthalpy change with certain structural parameters, especially changes in solvent accessible surface areas, have considerable predictive capability. We provide a physical foundation to that parametrization and in the process we present a system of equations that explicitly includes the enthalpic effects of the packing density between the different atoms within the protein molecule. The error in the prediction of folding/unfolding enthalpies at 60 C, is better than + 3 % (standard deviation = 4 kcal/mol).